Two tech giants launched new AI models today, India's sovereign AI program faces a compute gap, and the EU's deepfake labeling deadline is exactly two months away.
Google's Gemini 3.5 Flash is officially live: faster and cheaper than anything before it
Google made Gemini 3.5 Flash generally available today through Google AI Studio, the Gemini API, and Android Studio. The company says it delivers frontier-level intelligence at four times the speed of comparable models, priced at $1.50 per million input tokens and $9 per million output tokens. It ships with a one-million-token context window, meaning it can read an entire book's worth of text in a single request.
Why it matters: Speed and cost are the two main reasons businesses have been slow to adopt AI fully. A model this fast at this price changes the maths for students and small organisations who couldn't afford earlier tiers.
Microsoft Build 2026: seven new AI models and a unified "Microsoft IQ" intelligence layer
Microsoft announced seven new AI models at its annual Build developer conference, including MAI-Code-1-Flash for software coding tasks, MAI Thinking One for complex multi-step reasoning, and new models for image generation and speech processing. Alongside the models, Microsoft launched Microsoft IQ (a unified intelligence layer spanning workplace tools, cloud services, and the web) designed to embed AI across every Microsoft product simultaneously.
Why it matters: Microsoft is no longer simply reselling OpenAI. Building its own model stack means more competition in the market, which tends to lead to more choice and lower prices for anyone using AI tools for work or study.
AI music startup Suno raises $400M at $5.4B valuation. Coding robots get $400M too.
Suno, which lets anyone create original music using text prompts, closed a $400M Series D round led by Bond at a valuation of $5.4 billion. In the same funding window, Generalist AI (which builds robots that can handle complex physical tasks) raised $400M led by Radical Ventures at a $2B valuation. AI-powered market intelligence platform AlphaSense also closed $350M in a round backed by Accenture Ventures and J.P. Morgan Asset Management.
Why it matters: AI funding is no longer just about chatbots. Music creation, physical robotics, and knowledge work tools are all attracting serious investment, a sign that investors expect AI to reshape creative and physical jobs, not just office tasks.
EU AI Act: 60 days until deepfake labeling becomes legally required across Europe
The EU AI Act's transparency rules (Article 50) come into force in August 2026, giving companies roughly two months to comply. Any AI-generated content, including synthetic video, audio, and images, must be machine-readable marked as artificially made. Deployers in professional contexts must label AI-generated text on matters of public interest. Non-compliance carries fines of up to €7.5 million or 1.5% of global annual turnover. The European Commission's final Code of Practice guidance is expected this month.
Why it matters: This isn't a proposal or a debate, it is law with a live countdown. Every media company, content creator, and business using generative AI in Europe needs a compliance plan in place before August.
India's sovereign AI push hits a reality check: funds approved, compute still missing
Business Standard reported this week that India's sovereign AI programme is hitting infrastructure and financing hurdles despite strong government intent. BharatGen, the IIT Bombay-led consortium building India's first homegrown multimodal AI, received ₹1,058 crore in government backing (the most allocated to any of the 12 funded organisations), but still lacks the GPU compute required to train its planned trillion-parameter model. India's AI strategy currently remains dependent on foreign hardware and overseas data centres.
Why it matters: The gap between India's AI ambitions and its infrastructure is narrowing, but slowly. For students and developers, BharatGen represents real career and research opportunities, but the timelines are longer than the headlines suggest.
OpenAI Codex is now inside Amazon Bedrock, enterprise AI coding just got easier to deploy
OpenAI's Codex coding model is now generally available through Amazon Bedrock, covering both commercial and US GovCloud cloud regions. Businesses can access Codex through Amazon's familiar cloud infrastructure, complete with AWS-native security, governance, and compliance controls, removing the need to build separate API integrations from scratch.
Why it matters: For companies already on AWS, this significantly lowers the barrier to embedding AI-assisted coding into products and internal tools. It also accelerates AI adoption in regulated industries like finance and government, where AWS is already trusted infrastructure.
Deepfake job candidates: 1 in 3 applicant profiles may already be fabricated
A Pindrop Security study analysed 300 job applicant profiles and found that over one-third were entirely fabricated, complete with AI-generated résumés and deepfake video interviews that passed initial screening. Gartner projects that by 2028, 1 in 4 job candidate profiles globally will be fake. Companies are only beginning to build screening processes for this. The same tools making it easier to apply for jobs are making it harder to trust who you are actually hiring.
Pre-ChatGPT startups under pressure + Google's Gemini Omni arrives next month
CNBC reports that a generation of startups built before ChatGPT is being disrupted or shut down as AI makes their core tools obsolete and investor funding flows toward infrastructure rather than applications. Separately, Google revealed Gemini Omni (a new multimodal model designed to create video, audio, and images from any type of input) currently used internally, with a broader developer rollout expected next month.